In my last post Crowdsourcing Implications for Medium and Large Businesses, I looked at some of the other uses of crowdsourcing beyond the more targeted applications that are being readily adopted by home-office entrepreneurs and small businesses. These included examples in the areas of open-innovation and customer engagement. Another crowdsourcing opportunity relates to the selective out-tasking of highly repetitive, high volume work to a distributed online, on-demand workforce.

Labor-on-demand sourced via online platforms is akin to cloud computing but with people in the cloud not machines. As with cloud computing, where you can access multiple computers to harness more processing power, you can tap into a scalable work force whose supply will always be greater than any individual need for resources. Work can be completed more rapidly with many working in parallel on the completion of the tasks. There is low latency, 24/7 availability, and pay as you go pricing.

The issues however with the adoption of labor-on-demand within medium and large enterprises though are numerous:

1. Do the benefits outway the risks?
2. What type of work lends itself to being out-tasked at a medium to large enterprise level?
3. Will the savings significantly outweigh the management and quality control costs?
4. Will it lead to a better result than doing it in-house and/or will it speed up delivery?

There are four basic types of labor-on-demand work groupings, as follows:

• Micro Tasks – Generally very high volume, extremely low pay rate per task, and heavily automated throughout the process.
• Macro Tasks - Generally high volume, low rate of pay, mostly automated throughout the process.
• Simple Projects - Low volume or single tasks with moderate rate of pay, often requiring some direct contact with the worker. 
• Complex Projects - Single project with high rate of pay, typically requiring a substantial amount of time and direct interaction with the worker.

Large enterprises and providers of outsourcing services have been heavily investing over many years in developing models that replace labor based FTE (Full Time Equivalent) models, which involve contracting based on a number of resources over a given period of time. Large enterprises want to pass the risk and operation responsibility to the provider and therefore large contracts support the aggregation of work across business function (i.e. Benefits administration within HR) or single or multiple business processes (i.e. procure-to-pay).

Dividing the workload into thousands of micro-tasks and assigning it to thousands of workers is not viable for the typical back office functions (i.e. finance and accounting) or vertically integrated processes (i.e. check processing) where there is an established, multi-billion dollar outsourcing market. However, specific tasks that meet certain criteria have the potential to be sourced via a crowdsourcing model.

The adoption of crowdsourcing platforms for simple and more complex projects is well underway but for the high volume out-tasking of micro-tasks and macro-tasks, other than that for largely one-off initiatives, the model has some way to go before it can be considered mainstream.

The type of micro-task and macro-task work that can be sourced via crowdsourcing platforms currently falls into a few, very specific, categories.

Typical categories and examples include:

• Data optimization (i.e. removing duplicated data, categorizing product and image data, identifying similar groupings, data verification, mining and harvesting data)
• Language Accessibility (i.e. transcription and translation) 
• Market intelligence (i.e. competitive benchmarking) 
• Content management (i.e. content seeding and content moderation)
• Performance optimization (i.e. relevance tracking, search optimization)

In order for large volumes of work to be crowdsourced through online, labor-on-demand platforms, the work has to ideally meet a number of attributes, which include the following:

• The task and expected outcome should be easy to define.
• If the task requires judgment, tasks that require objective rather than subjective reasoning are better suited.
• It works well when you can establish quality metrics so that the checking of work quality can be automated for high volume tasks and simplified for lower volume tasks.
• The process or activity being crowdsourced should consist of a relatively small number of steps, ideally less than five, so it can be easily structured into a simple workflow.
• The work should be repetitive and there should be the need to perform a large number of tasks.
• The output or product should be easily re-configured with other tasks or seamlessly integrated into a larger process in order to produce a clean result.
  
  

Best illustrated by example, following are some user cases of companies and micro-task work that has been crowdsourced:

PayPal, an online payment system and a subsidiary of the US company eBay asked workers to research of information about “PayPal payment methods” for thousands of eBay entries. The information gathered by the workers gave PayPal invaluable data that enabled PayPal to learn about users preferred payment methods. 

IDG Business Media GmbH, the German subsidiary of leading global IT publisher IDG which publishes, among others, COMPUTERWOCHE, CIO and TECCHANNEL, asked workers to enter search engine optimized glossary texts for CFOworld.de, which had high search engine relevance, and were mainly financial sector related. Also, workers were asked to generate so-called product capsules for our TECCHANNEL.de platform involving the gathering of references, tests and reports about products or product categories from available free web content. The biggest challenge of this job was that the information had to be entered into the system in a highly standardized format so that it could then be entered automatically into the Content Management System.

Swink!TV – an American Internet media network for high school sports. Micro-task workers were asked categorize 3,700 videos of various high school sport events in order to identify the sport, home team, special plays, tag the images with pre-determined keywords and categorize them according to who was playing.

1stMOVER, an incubator for start-up companies that develop and market innovative products and services on the web asked workers to further advance their search engine/community at pinbooks.de by undertaking many jobs simultaneously including product tests, community seeding and market studies.

MSW & Partner is a human resources consulting firm that specializes in recruiting highly qualified college graduates, student teachers and young professionals. Workers were engaged to research of 8,500 department chairs and professors.


RentCycle needed to categorize more than 20,000 rental business websites in less than three weeks and presented the challenge to micro-task workers. Once the solution was designed, the online workforce sorted the sites accurately in under a week at a fraction of the cost of what it would have cost to hire someone to do it in-house.

University of Southern California initiated a project to reconnect with "lost alumni" - those graduates for which the school does not have updated contact information. Given the alums name and year of graduation, workers are finding Facebook pages, LinkedIn profiles, and Google references. The University will use this information to unobtrusively reach out and invite these alumni to join an online group where they can reconnect with their alma mater and fellow graduates.

A second initiative involves helping USC's storied athletic program reconnect with former athletes. USC's only comprehensive athletics records for most student athletes are in the original media guides and printed rosters for each sport. Workers are entering data from scanned images of the original rosters and media guides to create a master database of nearly every athlete to play at USC for more than 50 years.

Lombardi Sports decided to create an online catalog they faced a significant challenge gathering the product images and detailed descriptions they'd need for thousands of sporting goods would otherwise have taken months of work and days of review and management.  By deciding to crowdsource the work, the team broke thetasks into two parts. The first step was to gather images and find product descriptions, with built-in reviews at each step. The second step was for Lombardi Sports to provide the manufacturer name and product model numbers while the workers captured pictures and extracted detailed descriptions and specs from manufacturer websites. The finished results were loaded into an e-commerce solution and turned into a live online catalog.

So, crowdsourcing offers a range of possibilities for small, medium and large size enterprises. Two factors will ultimately indicate how big this is going to get – it is clear crowdsourcing is not a fad, but will it go mainstream and scale? 

Firstly, if that is going to happen, crowdsourcing either needs to create delivery mechanisms to fulfill new demand – work that previously could not be outsourced – or it needs to evolve and capture market share - the simpler tasks that are already outsourced. 

A second factor will be the extent to which crowdsourcing’s customers use crowdsourcing to source day-to-day demand on a recurring basis and Crowdsource repetitive tasks every day or every month rather than crowdsourcing as a series of one off projects, which is largely the case today.